The $1.20 SEO Stack That Replaces Your $200 Subscription

You were not paying for data. You were paying rent on a spreadsheet.

It is 8 a.m., and Marcus just cancelled his Semrush subscription

Marcus runs a seven-person landscaping company on the edge of Manchester. Every Monday, before his crew even loaded the vans, he opened Semrush to check three things: had a competitor outranked him for "garden landscaping quotes near me," had his site picked up any technical errors, and was the blog post his niece wrote last spring still bringing in traffic.

He used maybe ten percent of what he was paying for. Keyword research he needed twice a month. A rank tracker for nine phrases. One site audit a quarter. The other ninety percent of the dashboard, the content ideation tools, the ad-spend estimators, the backlink gap analysis for competitors three times his size, sat there unopened.

He paid for it anyway. Because the alternative was checking nothing, and checking nothing felt worse than overpaying.

That trade-off just disappeared.

Is there a free, open source alternative to Semrush and Ahrefs?

As of this week, yes, and it is spreading fast. A developer who goes by Ben on X posted two days ago: "Introducing OpenSEO, the open source alternative to Semrush and Ahrefs. I started building it out of spite because the existing tools were too expensive, bloated or scammy." The post has pulled in thousands of likes and bookmarks in under 48 hours, which in the AI-agent corner of the internet is the sound of something real hitting a nerve.

We pulled the repository ourselves this morning. OpenSEO covers the workflows a small business actually uses: keyword research, rank tracking, competitor insights, backlink checks, site audits, and a newer category called AI Visibility, tracking whether your business shows up when someone asks ChatGPT or Google's AI Overview a buying question instead of typing it into a search bar.

The code is free. What is not free is the data underneath it. OpenSEO does not maintain its own crawl of the internet, which is the expensive part Semrush and Ahrefs charge you for. Instead it connects to DataForSEO, an industry data provider, and you bring your own key. You pay DataForSEO directly, for exactly what you query, nothing more.

Where the AI agent actually fits in

Here is the part that makes this a Notma story, not a developer story. OpenSEO ships with an MCP server. MCP, short for Model Context Protocol, is the connector Anthropic built so an AI agent like Claude Code, or Codex, or Claude Desktop, can reach into a piece of software and use it directly, the same way you would, except it does not get bored on page four of a backlink report.

Point Claude Code at your OpenSEO instance and it stops being a chatbot that talks about SEO. It becomes an agent that runs the SEO. Ask it in plain language, "check where we rank against the two landscapers who show up above us for 'garden landscaping Manchester,' and flag anything broken on our site," and it queries OpenSEO, reads the results, and hands you an answer instead of a dashboard you have to interpret yourself.

That is the actual shift. Semrush gives you data and expects you to do the thinking. OpenSEO plus an agent gives you a finished judgment.

Why this is happening now, not three years ago

Two things had to be true at the same time. First, an API-only data provider like DataForSEO had to be cheap and good enough that nobody needed to build and maintain their own web index, which used to be the moat that let Semrush and Ahrefs charge what they charge. Second, AI agents had to get good enough, and cheap enough, to actually operate a tool instead of just answering questions about it.

Both landed this year. That is why a solo developer with a grudge about pricing could ship a real Semrush competitor as a side project, and why it has already picked up meaningful traction. The GitHub page showed roughly 4,000 stars when Ben posted on Saturday. We checked again this morning. It is past 5,100. That is over a thousand new stars in two days, which is not a fluke, it is a market telling you something.

Doesn't "self-hosted" mean I need to be a developer?

Less than you would think, and the project's own documentation says so directly: if a terminal command confuses you, copy it into ChatGPT or Claude and ask for an explanation.

There are two ways to run it. Docker is the fast path, meant for one person on one machine, and it is genuinely a few commands. Cloudflare is the path if you want it reachable from anywhere, or if more than one person on your team needs to log in, and the project's own instructions put that setup at five to ten minutes on Cloudflare's free tier. If you would rather not touch a server at all, OpenSEO also offers a hosted version at openseo.so, free to try, ten dollars a month if you want the paid tier, which is still a fraction of a legacy SEO subscription.

How to actually set this up, in one afternoon

1. Get a DataForSEO API key. Sign up at DataForSEO's site, request API credentials, and keep the login and password handy. This is the only account you are paying for going forward.
2. Self-host OpenSEO. For a solo test run: clone the repository, copy the example environment file, drop your DataForSEO key into it, and run docker compose up -d. It opens on your machine at port 3001. For a team, deploy the Cloudflare Worker version instead, which stays free on Cloudflare's free plan.
3. Connect your AI agent over MCP. In OpenSEO's dashboard, open AI & MCP and copy the server URL, then point Claude Code, Claude Desktop, or Codex at it with one line in your terminal.
4. Install the Agent Skill. OpenSEO ships pre-built Agent Skills for Claude Code that teach the agent how to run a keyword gap analysis, a competitor comparison, or a site audit without you writing the prompt from scratch each time. Before you install any third-party skill into an agent that touches your business data, it is worth reading what we wrote about vetting AI skills before you trust them with anything sensitive.
5. Ask for the finished job, not the raw data. Do not ask "show me keyword volumes." Ask "which of our top ten target keywords did we lose ground on this week, and why." That is the question a $200-a-month tool answers with a chart. This setup answers it with a sentence.

What Monday morning looks like now

Marcus did not build a dashboard habit. He built a five-minute conversation. Every Monday, he opens Claude Code and asks one question: "What changed in our search visibility last week, and is anything broken?"

The agent checks OpenSEO's rank tracker against his nine target phrases. It runs a quick site audit through the same connection. It compares his domain against the two competitors he actually cares about, not a generic industry benchmark. Then it writes him three lines: one keyword slipped two spots because a competitor added a new services page, one image on his quote-request page is loading slowly enough to matter, and traffic on the blog post about drought-resistant planting is up eighteen percent month over month.

He reads it over coffee. He has not opened a Semrush login screen in three weeks.

What you have after this

A week in, you have cancelled a subscription and stood up something that costs less than your morning coffee habit.

A month in, you have a Monday routine that takes five minutes instead of forty, because the agent already did the reading.

Six months in, you are not the person who checks rankings anymore. You are the person who gets told when something needs your attention, and otherwise gets left alone to run the actual business.

Same visibility. Ninety percent less rent.